Modelling of the filter-adsorber type air cleaner by using neural network

It is well known that most air purifying methods imply the passing of air flow, as a pollutant carrier, through a control unit which retains impurities. Properties of the air control unit and the purifying process itself therefore differ depending on the nature of present impurities, as well as on flow-thermal properties of air as the carrier of those impurities. For the assumed conditions, in terms of production of a pollution source and presence of different polluting substances in the form of dust, aerosols, gas, vapor in the exhaust gas, etc., an integrated gas purifier has been designed and tested, comprising a module for purification of mechanical impurities and a module for purification of gaseous impurities. The purifier is compact and has a universal application while simultaneously retaining several different pollutants. These requirements were met through application of the filtration and adsorption methods. On the formed experimental line with an adequate system of acquisition, filter-adsorber type gas cleaners in the function of flow-thermal parameters of gas mixture were tested simultaneously. Experimental data were used for training the radial basis function neural network, which was then used to model properties of the process and gas cleaner.